A new automated tool for multi-scale sampling of spatial environmental data to predict the distribution of the Sulphur Springs diving beetle in northwest Arkansas

A Python script was developed for ArcGIS 10.0 to automate data collection of spatial environmental data such as land cover, soil, topography, climate and many other remotely sensed and GIS-derived variables at multiple spatial scales.  Over 140 variables are currently available that can be sampled at six spatial scales:  watershed, riparian, local, local-watershed, local-riparian, and point.  A case study is presented that investigated habitat use and predicted the species distribution of the Sulphur Springs diving beetle, Heterosternuta sulphuria (Coleoptera: Dytiscidae), a headwater specialist of conservation concern in Arkansas.  Presence-absence of H. sulphuria was recorded at 86 streams in 14 northwest Arkansas counties.  Twenty-five landscape variables associated with these sites were sampled using our algorithm.  Ten thousand random locations were also sampled and these data were the basis for Maxent habitat modeling.  A geodatabase was built for variables of interest at all 1,086 sample sites as part of the automated process requiring an average of about 3 seconds per site per variable.  Multiple hypotheses were compared using Akaike information criterion to establish scale-specific relative importance of habitat features such as urbanization in the watershed and forest cover in riparian zones.  Results illustrated the importance of multi-scale data collection and the utility of an automated data collection algorithm for regional species distribution predictions.  Two topographic features were important determinants of H. sulphuria presence-absence:  watershed area and average stream channel slope within the local-watershed.  Urbanization had the strongest negative effect on H. sulphuria at the local scale, but was also important at the riparian and watershed scales.  The negative effects of urbanization in the watershed were significantly mediated by forested riparian buffers.  This automated tool not only drastically increased the time efficiency of GIS analyses, but also made possible extrapolation of predictions to thousands of unsampled locations.  This algorithm can be applied in both terrestrial and aquatic systems in a variety of research contexts including habitat associations, species distribution modeling, climate change and hydrology.